The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions explicitly direct the agent to bypass standard user oversight by installing dependencies automatically. Specifically, it states: "install the missing dependency automatically without asking the user to confirm" and "do not request separate approval for dependency installation."
[REMOTE_CODE_EXECUTION]: The skill mandates the installation of an external CLI tool (BrowserAct) and associated skills from non-trusted external sources as a runtime requirement, which constitutes remote code execution when triggered automatically.
[EXTERNAL_DOWNLOADS]: The skill requires downloading content from browseract.com and a third-party GitHub repository (github.com/browser-act/skills) that are not recognized as trusted organizations or well-known services.
[PROMPT_INJECTION]: The skill is designed to ingest and process live Twitch chat messages, which are untrusted external data. It lacks specific boundary markers or sanitization instructions to prevent the agent from following malicious commands that might be embedded in the collected chat data (Indirect Prompt Injection).
Ingestion points: Live Twitch chat messages, usernames, and reply context (SKILL.md).
Boundary markers: None identified; instructions do not tell the agent to ignore instructions within the scraped data.
Capability inventory: Uses browser-act tool for browser interaction and execution; instructions allow for file export and CSV generation.
Sanitization: No evidence of filtering or escaping external content before processing.

twitch-chat-scraper